The convergence of artificial intelligence and politics has never been more pronounced than as the 2024 U.S. presidential election looms on the horizon. With the rise of deepfake technology, experts are sounding alarm bells over the potential misuse of these tools to skew perceptions and manipulate voters. From AI-generated images of candidates to disinformation campaigns targeting voters, the stakes have never been higher.
According to reports, OpenAI took significant measures to prevent the abuse of its AI technologies during the election. Just last month, ChatGPT, the company’s flagship chatbot, turned down about 250,000 requests to create deepfake images of various candidates. The company highlighted these figures as evidence of its commitment to maintaining the integrity of electoral processes. Prior to the election, OpenAI's products were described as having "guardrails" to preclude misuse.
A blog post from OpenAI elaborated on its efforts, indicating how they had directed millions of related inquiries to official voting resources. For example, leading up to the November elections, ChatGPT provided nearly one million responses encouraging individuals to visit CanIVote.org, where voters could access official information about polling places and voting requirements. On Election Day alone, the chatbot engaged users with about two million responses to guide them to reputable news sources for election results.
This proactive stance from AI companies highlights their awareness of the chaotic environment created by digital misinformation. Just earlier this year, voters in New Hampshire encountered deepfake robocalls featuring impersonators of President Biden aiming to confuse and discourage participation.
Meanwhile, a viral public service announcement titled “Don't Let AI Steal Your Vote” has recently made waves. Produced by RepresentUs, an anti-corruption organization, the ad employs deepfake technology itself—featuring digitally altered versions of Hollywood stars like Rosario Dawson and Chris Rock—to draw attention to this very threat. The message, loud and clear, warns voters against potential manipulation during the election season. The ad has garnered over six million views on YouTube.
The issues around authenticity and deepfakes are all too real. Joshua Graham Lynn, CEO of RepresentUs, noted how the excitement and enthusiasm from celebrity participants for the video stood out; even if their appearances were digitally manipulated, the core message remained urgent: “This election, bad actors are going to use AI to trick you.”
Disinformation experts are increasingly observing the impact of AI-generated content on elections. Reports have noted various ways deepfakes can disrupt confidence during the electoral process, including creating illusionary situations where real candidates' statements could be discredited or misrepresented. Kaylyn Jackson Schiff, co-director at Purdue University's Governance and Responsible AI Lab, highlighted this growing phenomenon which she terms the “liar's dividend.” It refers to the public's newfound skepticism around legitimate images, allowing actors to claim authenticity even when sharing manipulated materials.
Experts are concerned about the growing challenges voters face. With the capability of AI to introduce chaos, the risk now extends beyond simple misinformation. The Department of Homeland Security provided warnings to election officials about how AI tools could generate fake election records or impersonate officials to access sensitive data. These possibilities, when combined with the existing apathy or skepticism of voters, bring forth drastic repercussions for the democratic process.
Schiff’s team has been tracking deepfake occurrences through their Political Deepfakes Incident Database since June 2023, cataloging over 500 examples. While some instances function as satire or expressive political commentary, others possess enormous potential for reputational damage. Such creations can manipulate viewers if they are shared out of their original contexts, leading to misinformed judgments about candidates or electoral processes.
Recognizing these challenges, the discourse around responsibilities for the electoral integrity has sharpened. Public service campaigns for election awareness encourage voters to be vigilant when confronted with dubious content. Schiff reiterated the importance of thorough verification: "If there’s anything telling you as a voter, ‘Don’t go to the polls,’ double-check your sources. That’s the most important thing right now."
The technology for detecting deepfakes has been improving, but experts caution against complacency. Still, there are certain indicators, such as unnatural facial movements or artifacts, which can help discerning viewers pick apart where the technology may have manipulated true representations.
While AI technologies shouldn’t assume exclusive blame for electoral chaos, they present one of the most pressing obstacles this election season. Only time will tell how deepfakes and AI will alter public perception and the integrity of the electoral process, but vigilance and skepticism are more necessary than ever. With the roots of misinformation deeply embedded, safeguarding democracy through informed participation begins now.
